| Industries that have important network features have long been a concern of public policy, since network industries often embody two major forms of potential market failure: significant economies of scale and externalities. I examine whether network externalities are indeed technological externalities, and the impact when these externalities are internalized by both consumers and firms. I begin by demonstrating that when there are increasing returns to scale in the production of software, indirect network externalities have the same form of technological externality—that is, an adoption externality—as direct network externalities. I then demonstrate that when the indirect network externality is internalized on the supply side, it does not impose the corresponding monopoly efficiency losses, since the monopoly hardware producer realizes the connection between the demand for hardware and the number of software varieties. Finally, simply internalizing the externality still leaves the difficult problem of coordination. I examine the extent to which rational individuals manage to coordinate their choices in a sequential choice framework. I approach the coordination problem in a framework suggested by the literature on information cascades. I use a dynamic programming approach that allows individuals to calculate a subjective probability distribution over the probability of the remaining individuals to choose one good or the other, dependent upon both the choices of previous individuals and upon her own choice. This approach allows me to characterize a coordination cascade. That is, I am able to determine the circumstances where individuals ignore their own preferences when making their choice, and simply make the same choice as everyone else. Building on these cascade conditions, I am able to characterize the probability distributions of the equilibria of the sequential choice process. I discover that when the network externality is large, a substantial amount of coordination is achieved, and although perfect coordination is never guaranteed, expected relative efficiency is high. |
